Applying an electric field to control metals in furnaces
08/01/2000
Daniel M. Dobkin,* Igor Rapoport, Vladimir Starov, Sizary Inc., Los Gatos, California**
Yossef Raskin, Solomon Zaidman, Sizary Ltd., Migdal Tefen, Israel**
*Currently with WJ Communications, Milpitas, California
**Additional authors are listed in the Acknowledgments
overview
Despite intuition, application of an electric field does have a measurable effect on the movement of metallic contaminants in gas in a wafer processing thermal process. Research has shown that the phenomenon of positive surface ionization can give rise to significant electric field effects on the gas-phase transport of ionizable atoms such as alkali metals, even at very low fractional ionizations and modest electric fields. In addition, such application of an electric field can also reduce contamination by transition metals at high temperatures.
 Figure 1. Detailed balance at a surface with neutral evaporation and small fractional ionized evaporation. |
Contamination in thermal processes such as oxidation, diffusion, or annealing can arise from various sources. For example, heating elements release volatile contaminants that can diffuse through quartz furnace tubes [1]. Silicon carbide furnace hardware is typically coated with a chemical vapor deposition (CVD) layer to prevent outdiffusion from the relatively dirty bulk ceramic, but may still be a source of iron and other metals. Once in a furnace, volatile metal atoms normally are transported by convection and diffusion in ambient gases. Some of these undesirable metals are incorporated into wafers being processed.
We have studied the effects of applying DC-electric fields to control the transport of metals and reduce or prevent contamination of wafers. We have found that moderate electric fields have significant effects on the concentrations of alkali and transition metals on a wafer surface and in bulk silicon, and can be used to reduce contamination of processed wafers effectively. These effects can be explained by examining the mechanisms of ionization and ion and neutral transport in silica and in gaseous media.
Field effects in the gas phase
At first glance, one would be inclined to assume that electric fields should have no measurable effect on metallic contaminants in gas. For example, it is easy to show that even for easily ionized metal atoms such as sodium with a first ionization energy of 5.1eV, the thermal equilibrium fractional ionization in the gas phase (i.e., the ratio of ionized to neutral sodium atoms) should be ~1:1010 at 1000°C. Since the effects of a small electric field on neutral atoms are quite negligible, this would at first lead us to conclude that electric fields should have no effect at all.
We must, however, account for interactions at surfaces. It has been known for many years that atoms can be ionized by electron-exchange processes during their impact upon a surface if the electron affinity (equivalently, the barrier height for electron emission) of the surface is comparable to the ionization energy of the atom [2]. This phenomenon—positive surface ionization—has been studied extensively for metal surfaces [3]. The fractional ionization (i.e., fi or the fraction of evaporating atoms from a surface that are ionized) can be approximately described by the Saha-Langmuir equation for fi <1:
 |
null
where Ø is the barrier height, Ei the ionization energy of the evaporating atom, k Boltzmann's constant, and T absolute temperature. For silicon, the electron affinity is about 4.05eV [4]. Thus the fractional ionization of sodium evaporating from a silicon surface around 1000°C is about 5x 10-5. This fractional ionization, while small, is greatly in excess of that obtained in the absence of surface processes.
 |
For more complex, covalent surfaces, it is more difficult to estimate fractional ionizations; local interaction of the departing atom with individual surface atoms must be accounted for [5]. One can nevertheless argue that the fractional ionization is certainly not decreased by the addition of more electronegative atoms, such as oxygen, in the case where ionization takes place at an oxidized silicon surface. Thus the Saha-Langmuir estimate represents at least a lower bound on the fractional ionization of evaporating alkali atoms from silicon or silicon dioxide surfaces.
The small fraction of atoms that ionize when striking a surface is enough to produce a very significant redistribution of atoms in the gas phase. The essential physics is the result of the very high drift velocity of ions under the influence of an electric field, relative to diffusion of neutrals. If we consider transport of ions across a gap (g = 1cm) under the influence of an electric field of 400V/cm at atmospheric pressure, using a typical mobility of 3cm2/v-sec [6], we find that ions move at a velocity of 1200cm/sec, with the transit time across the gap being about 0.8msec. The corresponding effective velocity for diffusion of neutrals is (D/g), where D is the binary diffusivity of neutral atoms in the gas; a reasonable value is about 1cm2/sec at atmospheric pressure and 1000°C, giving diffusion velocities on the order of 1cm/s, or three orders of magnitude slower than ion-drift velocities. Very small concentrations of ions can lead to large fluxes, requiring significant changes in neutral concentration to compensate them.
We first estimate the concentration of ionized atoms in the near-surface region for a given fi <1 by assuming that the atoms in the silicon dioxide are present as an ideal solution so that the number of sodium atoms, for example, evaporating per unit area per unit time is simply proportional to the near-surface concentration of sodium. We also assume that detailed balance would hold for pure sodium; that is, the number of atoms evaporating in equilibrium from a pure sodium surface would equal the atoms striking the surface from a sodium vapor at the equilibrium vapor pressure, which we obtain from kinetic theory as noc/4, where c is the thermal velocity—(8kT/pm)1/2—and no the concentration of neutral atoms (Fig. 1). We obtain the ratio of ionized to neutral atoms near the surface as
 |
null
where µ is ion mobility, Eo electric field, and m atom mass.
We can then assert that in steady state the diffusion and drift fluxes must balance each other in the gas phase. This is valid for short times, since fluxes in the gas phase greatly exceed those in solid boundaries (e.g., a silicon wafer or furnace tube). We find the interesting result that the electric field value cancels out of the expressions because increasing electric field causes a decrease in ion concentration from Eqn. 2 that is exactly balanced by the increase in ion velocity, resulting in an ion flux independent of field. The concentration of neutral atoms must fall near the positive wall to create a diffusion flux balancing the drift of ions. We obtain an expression relating the concentration of sodium at the surface (nsurf) with and without an electric field applied:
 |
null
where g is the gap between the two walls, and D binary diffusivity.
In Table 1 we show the effect of electric field, as the ratio of surface neutral concentration with and without the field, for a range of fractional ionization values. We have also included the temperature required to produce such fractional ionizations from a silicon surface, using the Saha-Langmuir equation for sodium atoms. We can see that an applied electric field should have a significant effect on the transport of sodium atoms from a silicon wafer at temperatures in excess of about 700°C.
Experimental results: alkali metals
To study the effects of the temperature dependence of electric fields, we intentionally contaminated test wafers with 100nm of oxide by implanting them with sodium at 100keV, to a dose of 1.2x 1013 atoms/cm2. We placed these next to uncontaminated wafers that were held at ground while 425V were applied to the test wafer (Fig. 2a). After annealing at various temperatures for 1 hr in argon, we measured the sodium in the samples with atomic adsorption spectrometry (AAS); the resulting concentrations are shown in Fig. 2.
Table 2 gives the ratio of concentration on the positive and grounded samples as a function of temperature. For comparison, we show the predicted magnitude of the field effect based on our model, using a thermal velocity of 105cm/sec and a 2cm gap, with fractional ionization predicted by the Saha-Langmuir equation for sodium on silicon.
We found very reasonable qualitative agreement between the model and the measured response, particularly considering the fact that we have made no attempt to incorporate the fairly complex effects of simple evaporative loss of sodium due to diffusion and convection in the gap region during anneal, or to account for the difference in surface ionization between clean silicon and the actual oxidized surface.
 |
In a second set of experiments, we varied the applied electric field. In these experiments, we placed a batch of wafers in a radiatively heated quartz-wall furnace with all wafers held at a positive potential with respect to an outer mesh electrode (Fig. 3). We annealed the oxidized silicon wafers at 1000°C in argon for 120 min, with various applied electric fields, and compared them to samples for which no electric field was used. The gap between the wafer edges and the outer electrode was roughly 2cm. We subsequently measured the concentration of sodium and lithium in the surface oxides by secondary ion mass spectroscopy (SIMS). Only the integrated alkali concentration (ions/cm2) is reported here, to avoid any artifacts due to the migration of ions within the oxide during the SIMS measurement. No intentional source of contamination was present.
Figure 4 shows measured sodium concentration vs. applied voltage. We observed that lithium behavior was very similar to that of sodium, as expected from their similar ionization energies.
We can see that application of a positive voltage to the wafers significantly reduces the level of alkali metals. The range of sodium concentration when the field was on is quite comparable to that of control wafers that received no processing at all; thus, we can tentatively interpret the results as showing that sporadic ambient contamination present in the furnace was effectively repelled by the field, preventing its incorporation into the wafers. The magnitude of the electric field, as measured by the applied voltage, has very little effect on the contaminant distribution function, just as was predicted by theory discussed above.
It thus appears that the phenomenon of positive surface ionization can give rise to significant electric field effects on the gas-phase transport of ionizable atoms such as alkali metals, even at very low fractional ionizations and modest electric fields.
Transition metals
 Figure 4. Measured sodium areal concentration by SIMS as a function of applied voltage. Also shown are control samples that received no anneal prior to measurement. |
When processing is performed at higher temperatures, 1050-1350°C, silicon carbide (SiC) tubes are often used instead of quartz. Unlike quartz, SiC is a good diffusion barrier even at these high temperatures, and transport of contaminants from sources outside the diffusion tube is greatly inhibited. Metallic contaminants (i.e., iron and other transition metals in particular) can, however, still be introduced inside the tube from other sources, such as contaminated gases and wafers. Each defective, improperly treated, or worn out SiC tube can itself be a source of contamination.
 Figure 5. Iron concentration with and without electric field. |
We have found that the application of an electric field can also reduce contamination by transition metals at high temperatures. In Fig. 5, for example, we show results for bulk iron concentration using µPCD [7], for wafers processed in a horizontal SiC tube with a SiC paddle and boats, at zero field or with 1000V applied to the wafers relative to the grounded tube. All the SiC parts were CVD coated and wet-cleaned prior to use, but the SiC was not oxidized or cleaned using Cl-containing gases. As a consequence, fairly high iron contamination levels are observed as temperature is increased. Application of the electric field leads to a significant reduction in iron contamination at temperatures >1100°C, though the effect is less dramatic than that observed for alkali metals.
By comparison to alkali metals, transition metals have higher ionization potentials (e.g., the first ionization energy of iron is 7.9eV vs. 5.1eV for sodium). To explain the observed effects, we speculate that surface ionization can still be significant if electronegative atoms such as oxygen are present on the surface, as a result of localized electron transfer.
Conclusion
We have found that moderate electric fields can be exploited to influence significantly the transport of ionizable impurities of commercial importance, such as alkali metals and iron, in the gaseous ambients typical of a semiconductor-processing environment. The electric-field induced reduction in metallic contaminants is probably related to surface ionization processes. These effects are robust, and show little dependence on the magnitude of the electric field.
We should note that as a consequence of these observations, it is possible to exploit electric fields imposed within a furnace environment to prevent transport of metallic contaminants from the furnace ambient to wafers in process; this has been described in US Patent 5,770,000. Furnace modifications to exploit these effects are available commercially.
Acknowledgments
The authors thank Vladimir Gorodokin for valuable contributions to this work.
References
- P.F. Schmidt, J. Electrochem. Soc., 130, 196, 1983.
- J. Taylor, I. Langmuir, Phys. Rev. 44 #6, p. 423, 1933.
- "Atomic and Ionic Impact Phenomena on Metal Surfaces," M. Kaminsky, Academic Press, New York, 1965.
- S. Sze, VLSI Technology, 2nd edition, McGraw-Hill, New York, p. 656, 1988.
- N. Winograd, B. Garrison, Low Energy Ion-Surface Interactions, ed. J. Rabalais, John Wiley & Sons, New York, p. 264, 1994.
- S. Brown, "Basic Data of Plasma Physics," AIP Press, Woodbury, p. 78, 1994.
- "Ultraclean Surface Processing of Silicon Wafers: Secrets of VLSI Manufacturing," ed. T. Hattori, Springer, Chapter 15, 1998.
Daniel M. Dobkin received his BS from Caltech and his MS and PhD from Stanford University. Previously with Sizary, Dobkins is director of technical marketing at WJ Communications, 1530 McCarthy Blvd., Milpitas CA 95035; ph 408/433-5663, e-mail [email protected].
Igor Rapoport received his MS in electronic engineering from Zaporozhye Technical University, Ukraine. Rapoport is customer applications group leader at Sizary Inc., 170 Knowles Dr., Suite 201, Los Gatos, CA 95032; ph 408/364-3815, fax 408/364-3811.
Vladimir Starov received his PhD in chemical physics from Brandeis University. Starov is president of Sizary Inc.
Yossef Raskin received his MSc in physics and electronics from Gorky State University, Russia. Raskin is process development group leader at Sizary.
Solomon Zaidman received his MSc degree in electronics from Radio Technical Institute, Ryazan, Russia. Zaidman is an applications engineer at Sizary.